The current south asian floods due to heavy rainfall and Texas flood due to hurricane Harvey has had a devastating affect on many people's lives. Some point to climate change, some disagree. I saw a you-tube video that says the tipping point has passed, true or not, weather extremities seems to be on the rise. I haven't looked at much of the datasets myself, but removing speckle noise, a form of multiplicative noise from radar imageries, will likely be of value in understanding these events.
Google Earth Engine combined with ESA's Sentinel missions is an incredibly useful utility to monitor environmental phenomenons. Unaffected by clouds, sentinel-1 radar imageries, in the form of GRD product, are suitable for analyzing water bodies and flood extents, specifically given that water falls into the lower end of the backscattering coefficient, and in sharp contrast to the surrounding.
Though Google earth engine has the basic API for dealing with filtering, reducing, analyzing and visualizing petabytes of raster and vector datasets, I could not find any built-in speckle noise removal routines, which is essential when dealing with any radar imageries. So I have implemented a simple algorithm from the paper "Enhanced Directional Smoothing Algorithm for Edge-Preserving Smoothing of Synthetic-Aperture Radar Images", in JavaScript using earth engine API. The loop in the EDS Algorithm in the paper, was implemented by convolving the image with directional kernels, and earth engine reducers to pick pixels with minimum values. All code runs on earth engine server, i.e no client side object. Mean squared difference as a measure for speckle reduction, as suggested in the paper, is significantly better with EDA as compared to plane smoothing.
In future I will probably generalize the code to use other kernel size, now it uses 3x3. And the kernel weights appropriations could be improved perhaps. The floods are happening so i thought i should release this in-case someone finds it useful and optimize later :-). If you have an earth engine account, you could test the implementation on their code-editor, the default area is a subset of Texas, and the image filtered on recent sentinel-1 acquisition.
Reference:
Mastriani, Mario, and Alberto E. Giraldez. "Enhanced directional smoothing algorithm for edge-preserving smoothing of synthetic-aperture radar images." arXiv preprint arXiv:1608.01993 (2016).
My Earth Engine Implementation
